The Longevity of the Relational Database

  • Comments posted to this topic are about the item The Longevity of the Relational Database

    Best wishes,
    Phil Factor

  • Just because the relational model has a very good track record, doesn't mean we shouldn't be careful what we take for granted. For example, if someone had told me even five years ago, that today I would be running more Linux machines in my home than Windows machines, I would have thought they were stark raving bonkers. Back then I had Linux (fedora) as a bit of a curiosity, dual booting on one machine along side XP. I had two other machines; a Vista and another XP. Windows outnumbered Linux 3:1, and my Linux install served no real purpose except mild amusement.

    Today my family has between us:

    :: 4 phones, all running Linux

    :: A Galaxy tab, running Linux

    :: A NAS, running Linux out of the box

    :: A Vista machine

    :: An XP machine (dual booting Ubuntu)

    :: A Tivo running Linux (I'm pretty sure it's Linux)

    :: An old box given a new lease of life running Ubuntu (still just a plaything).

    Linux outnumbers Windows 9:2 and it just... happened. There was no conscious decision for much of this. I learned the other day than a large employer near here (13,000 EFT) is dropping MS Office for Google docs. Management's attitude is "the staff will just have to adapt". I wish them luck, sounds like my idea of hell.

    The point is that these changes can sneak up on us, and presumably data management is not immune. The biggest user of organised data in the world would have to be Google. Is all/any of their data organised in the conventional relational way? Just as Windows missed (perhaps intentionally) the evolution of the OS into almost a disposable consumer item, RDBMSs could go the same way. We may not notice it at first because the RDBMSs reach may be still expanding, but if they fail to expand as rapidly as the "next big thing" they may eventually fade into the niche world.

    ...One of the symptoms of an approaching nervous breakdown is the belief that ones work is terribly important.... Bertrand Russell

  • Phil,

    Can you name some of these 'new generation of databases'/'NoSQL' products/attempts/failures.

    Thanks.

  • Good reading, thanks. The NoSQL movement is more than 20 years old? Didn't know that...

    Hakim Ali
    www.sqlzen.com

  • Michael Meierruth (9/17/2012)


    Phil,

    Can you name some of these 'new generation of databases'/'NoSQL' products/attempts/failures.

    Thanks.

    Google's BigTable

    Facebook's Cassandra

    MongoDB

    JSON is seen as a bit of a data store, I guess XML as well

    Neo4J and other graph databases.

    There's also Streambase and streaming technologies that displace RDBMSes in some ways.

  • GPO (9/16/2012)


    Just because the relational model has a very good track record, doesn't mean we shouldn't be careful what we take for granted.

    ...

    Good points, though I'm not sure the OS analogy is a great one. Many of what you mention are new types of devices that consume other services. The RDBMS still exists, though the big 3 (Oracle/DB2/SQL Server) are losing some of their dominance. The SQLLite, MySQL, etc. products have grown, but these are still RDBMS products.

    There is likely to be change over time that grows other types of databases, and that makes sense. Some domains of problems need different solutions. The RDBMS works, but it's being shoehorned in places where something else does better. I think that we will always have RDBMS, and after file stores, these will have lots of data, but we'll see other types of stores (document, streaming, graph, maybe more) over time.

  • Phil,

    Can you name some of these 'new generation of databases'/'NoSQL' products/attempts/failures.

    Thanks.

    I've just had a rummage around, but I haven't yet found my notes from the early nineties so I can't give you a full list. (I had to give up after sneezing at the dust!) However, the page Object Database should be a good starting point. I was going to give you a couple of examples that I could remember, but it seemed unfair to single them out since the people who worked so hard on them are still in the industry, probably, and my points were general, rather than aimed at the two whose names I can remember! (Also, one of them has actually limped on to the present day, evidently after being bought a few times) Nobody had invented the 'NoSQL' umbrella term then.

    Best wishes,
    Phil Factor

  • There are plenty of places where absolute consistency is less important than size and speed. Many times I think a relational system is overkill for those applications.

    But to expect that the relational model (or some very close variant) will no longer be required is like saying that arithmetic is obsolete.

    ...

    -- FORTRAN manual for Xerox Computers --

  • I actually have to keep asking developers "Why do you want to put that in SQL Server? What value does SQL Server and a relational model add to that data? What are you going to join it to? What does it relate to? Did you consider the licensing costs?"

    SQL, as a mostly-relational database, is good for some things. It is, however, a huge amount of overhead for other things (write only logs that don't relate to anything, for instance). It's even very bad at some things (properly constrained temporal databases), though sometimes there isn't a better alternative available. Sometimes there is.

  • GPO (9/16/2012)


    Just because the relational model has a very good track record, doesn't mean we shouldn't be careful what we take for granted. For example, if someone had told me even five years ago, that today I would be running more Linux machines in my home than Windows machines, I would have thought they were stark raving bonkers. Back then I had Linux (fedora) as a bit of a curiosity, dual booting on one machine along side XP. I had two other machines; a Vista and another XP. Windows outnumbered Linux 3:1, and my Linux install served no real purpose except mild amusement.

    Today my family has between us:

    :: 4 phones, all running Linux

    :: A Galaxy tab, running Linux

    :: A NAS, running Linux out of the box

    :: A Vista machine

    :: An XP machine (dual booting Ubuntu)

    :: A Tivo running Linux (I'm pretty sure it's Linux)

    :: An old box given a new lease of life running Ubuntu (still just a plaything).

    Linux outnumbers Windows 9:2 and it just... happened. There was no conscious decision for much of this. I learned the other day than a large employer near here (13,000 EFT) is dropping MS Office for Google docs. Management's attitude is "the staff will just have to adapt". I wish them luck, sounds like my idea of hell.

    The point is that these changes can sneak up on us, and presumably data management is not immune. The biggest user of organised data in the world would have to be Google. Is all/any of their data organised in the conventional relational way? Just as Windows missed (perhaps intentionally) the evolution of the OS into almost a disposable consumer item, RDBMSs could go the same way. We may not notice it at first because the RDBMSs reach may be still expanding, but if they fail to expand as rapidly as the "next big thing" they may eventually fade into the niche world.

    And even the Win XP/Vista/7 machines have bits of BSD Unix in the kernel and its periphery. Not Linux, but *nix nonetheless.

    - Gus "GSquared", RSVP, OODA, MAP, NMVP, FAQ, SAT, SQL, DNA, RNA, UOI, IOU, AM, PM, AD, BC, BCE, USA, UN, CF, ROFL, LOL, ETC
    Property of The Thread

    "Nobody knows the age of the human race, but everyone agrees it's old enough to know better." - Anon

  • hakim.ali (9/17/2012)


    Good reading, thanks. The NoSQL movement is more than 20 years old? Didn't know that...

    I think NoSQL is a new name for a much older concept.

    In fact I think the underlying concept predates relational theory.

    Some of the "new" systems look remarkably similar to COBOL flat files.:hehe:

  • David.Poole (9/18/2012)


    hakim.ali (9/17/2012)


    Good reading, thanks. The NoSQL movement is more than 20 years old? Didn't know that...

    I think NoSQL is a new name for a much older concept.

    In fact I think the underlying concept predates relational theory.

    Some of the "new" systems look remarkably similar to COBOL flat files.:hehe:

    Everything old is new again.

    "Cloud Computing" is just "dumb-client" born again. 1970s technology with a better marketing-name.

    "NoSQL" is just hierarchical, pointer, or flat-file databases born again, with a retronym added for marketing (again).

    And so on.

    - Gus "GSquared", RSVP, OODA, MAP, NMVP, FAQ, SAT, SQL, DNA, RNA, UOI, IOU, AM, PM, AD, BC, BCE, USA, UN, CF, ROFL, LOL, ETC
    Property of The Thread

    "Nobody knows the age of the human race, but everyone agrees it's old enough to know better." - Anon

  • As usual, great article Phil. Thanks!

  • GSquared (9/18/2012)


    Everything old is new again.

    "Cloud Computing" is just "dumb-client" born again. 1970s technology with a better marketing-name.

    "NoSQL" is just hierarchical, pointer, or flat-file databases born again, with a retronym added for marketing (again).

    And so on.

    Yes and no. The basic idea is the same, but the way cloud computing paritions dynamically and the scales available are new. The fact you don't need everything on the server, and substantial amounts of work can be done on the client with Javascript, are new.

    NoSQL is similar, but they are again adding to the concept and making it a much stronger offering in some domains. The scalability and the ways in which there is knowledge of the physical location, multiple copies, etc. are substantial additions, IMHO.

Viewing 14 posts - 1 through 13 (of 13 total)

You must be logged in to reply to this topic. Login to reply